[Skip to Content]
[Skip to Content Landing]
Comment & Response
December 19, 2017

Benefits and Risks of Machine Learning Decision Support Systems

Author Affiliations
  • 1Department of Radiology, Milton S. Hershey Medical Center, Hershey, Pennsylvania
JAMA. 2017;318(23):2355-2356. doi:10.1001/jama.2017.16611

To the Editor I disagree with the premises and conclusions of the Viewpoint by Dr Cabitza and colleagues on the unintended consequences of machine learning in medicine.1

The authors’ conclusion raises the bar for artificial intelligence higher than that for new pharmaceuticals, medical devices, or changes in care delivery. Cabitza and colleagues espoused a standard that artificial intelligence must provably affect clinically important and relevant outcomes, as well as satisfy patients and physicians. Yet imaging artificial intelligence tools have already received regulatory approval this year in Europe, Australia, and New Zealand. Well-designed artificial intelligence algorithms exist today for the detection of pulmonary tuberculosis,2 breast malignancy,3 and serious brain findings such as stroke, hemorrhage, and mass effects.4